Parental Smoking Cessation and Adolescent Smoking

Journal of Pediatric Psychology, Vol. 27, No. 6, 2002, pp. 485–496

Parental Smoking Cessation and Adolescent Smoking Laurie Chassin,1 PhD, Clark Presson,1 PhD, Jennifer Rose,2 PhD, Steven J. Sherman,2 PhD, and Justin Prost1 Arizona State University and 2Indiana University

1

Objective: To examine the relation of parent smoking cessation to adolescent smoking and test its potential mediators. Method: Participants were 446 adolescents and their parents who completed a computerized measure of implicit attitudes toward smoking and questionnaires assessing smoking, parenting, and explicit attitudes. Results: Parental smoking cessation was associated with less adolescent smoking, except when the other parent currently smoked. In general, ex-smoking parents showed more antismoking socialization than did smoking parents. However, in children’s reports, these effects were negated if the other parent (particularly the mother) smoked. Children’s reports of parents’ antismoking behavior partially mediated the relation between parental smoking and adolescent smoking. Although children’s implicit and explicit attitudes were unrelated to parental smoking, mothers’ implicit attitudes were related to both their own smoking and their child’s smoking. Conclusions: Parental smoking cessation may help lower risk for adolescent smoking. However, this benefit may be realized only if the other parent does not currently smoke. Antismoking parenting might be a useful focus in cessation interventions. Key words: parent smoking cessation; parenting; adolescent smoking. The role of parent smoking in adolescent smoking has been somewhat controversial, with some researchers viewing it as relatively unimportant and others finding it a robust predictor (Conrad, Flay, & Hill, 1992). The inconsistency may stem from differing definitions of parent smoking (i.e., current vs. lifetime smoking; Bauman, Foshee, Linzer, & Koch, 1990), child smoking (e.g., experimental smoking vs. persistent smoking), and differing demographics (e.g., parental smoking may be more influential for non-Hispanic Caucasians than for others, U.S. DeAll correspondence should be sent to Laurie Chassin, Psychology Department, Box 871104, Arizona State University, Tempe, Arizona 85287–1104. E-mail: [email protected].

© 2002 Society of Pediatric Psychology

partment of Health and Human Services [USDHH], 1994). Despite inconsistent findings, there are several reasons to focus on the role of parent smoking. First, parent smoking is related to the persistence of adolescent smoking (Flay, Hu, & Richardson, 2000) and to smoking trajectories that are particularly problematic because they show early onset, rapid escalation, and long-term persistence (Chassin, Presson, Pitts, & Sherman, 2000). Second, because tobacco dependence shows significant heritability, parental smoking may reflect important genetic influence (Heath & Madden, 1995). Psychosocial research has often conceptualized the role of parent smoking in terms of social learning

486

theory, hypothesizing that modeling and access to cigarettes raise adolescents’ risk to smoke (Flay, Petratis, & Hu, 1999; USDHHS, 1994). These theories focus on the deleterious effects of adolescents’ exposure to parents’ current smoking, and parents who have quit are simply viewed as nonsmokers. Conversely, behavioral genetic theories are more likely to consider parents’ lifetime smoking, because those who have ever been regular smokers might transmit a genetic predisposition that raises their offspring’s risk to smoke (e.g., Boomsma, Koopmanns, van Doornen, & Orlebeke, 1994). However, neither of these approaches focuses on the potential importance of parental cessation. A few studies have found that parent smoking cessation is associated with less adolescent smoking. Jackson and Henricksen (1997) reported that children of two ex-smokers had less smoking onset than did children of two current smokers, but higher onset rates than children of two nonsmokers. Farkas, Distefan, Choi, Gilpin, and Pierce (1999) found that parental smoking cessation had the strongest effect if it occurred when the child was younger than nine years old. Most important, Bricker et al. (2001) found that parent smoking cessation (when the child was in third grade) prospectively predicted lowered risk of adolescent smoking 9 years later. However, none of these studies examined mediating mechanisms for these effects (although Bailey, Ennett, and Ringwalt [1993] found that general parenting styles did not explain parental smoking cessation effects). This study is the first to examine smoking-specific parenting and attitudes that might mediate the relation between parent smoking cessation and adolescent smoking. Theoretically, parent smoking cessation might influence adolescent smoking by several mechanisms. Social learning theory suggests that home environments without adult smoking models will produce lower risk for adolescent smoking. In addition, parents who have quit smoking might provide particularly strong antismoking socialization to their children. They may be less likely to tolerate smoking by their adolescents (perhaps to even a greater degree than nonsmokers, given their own struggles to quit). A large literature has shown that adolescents who view their parents as particularly intolerant of their smoking are less likely to smoke (e.g., Chassin, Presson, Sherman, Corty, & Olshavksy, 1984; see Flay et al., 1999, for a review). Moreover, parents who have quit smoking may be particularly motivated (even more so than nonsmokers) to protect their chil-

Chassin, Presson, Rose, Sherman, and Prost

dren from smoking initiation. Thus, their parenting behaviors may include punishing smoking and discussing smoking with their children. These behaviors are associated with lowered likelihood of adolescent smoking (Chassin, Presson, Todd, Rose, & Sherman, 1998; Jackson & Henricksen, 1997). However, in other ways, parental smoking cessation might actually elevate the likelihood of adolescent smoking compared to parents who never smoked. Ex-smoking parents may not perceive themselves as having legitimate authority to regulate their child’s smoking behavior because they do not feel comfortable demanding that their children behave in ways that they themselves did not. In turn, their children may not perceive them as having the legitimate authority to regulate their smoking. Research on parenting suggests that such a perceived lack of legitimacy reduces parents’ efforts to engage in socialization practices and also undermines the success of their efforts (Grusec & Goodnow, 1994). In addition to parenting factors, attitudes toward smoking might also explain the effects of parent smoking cessation. For example, because of their struggles to quit, parents who have stopped smoking may have particularly negative attitudes toward smoking, and they might communicate these negative attitudes to their children. A large literature suggests that negative attitudes toward smoking prospectively predict low rates of smoking behavior (see Flay et al., 1999, for a review). However, attitudes toward smoking may also be associated with heightened risk for adolescents with ex-smoking parents. The fact that parents did smoke at some time might inadvertently communicate to their children that smoking has some positive benefits. If there were no positive effects of smoking, why would a parent have been a smoker for any length of time? Thus, children of ex-smoking parents may view smoking as relatively positive compared to children whose parents have never smoked, and this could increase their risk for smoking. Previous studies that have examined the relations between attitudes and smoking behavior have relied on explicit (pencil and paper) measures of attitudes, which are under conscious control and are influenced by social desirability. Although these explicit measures are good predictors of smoking behavior, they may be limited in their ability to predict the subtle messages that parents transmit to their children, because these messages are likely to be beyond the parents’ awareness. In fact, explicit attitudes toward smoking have not shown transmission from

Parental Smoking Cessation

parent to adolescent (Chassin, Presson, Rose, & Sherman, 1998). Research in social psychology suggests that explicit measures are good predictors of behavior that is under conscious control but not behavior that is beyond awareness (Greenwald, McGhee, & Schwartz, 1998), which may be better predicted by implicit measures. Thus, in predicting adolescents’ smoking behavior from their parents’ attitudes, the parents’ implicit attitudes rather than their explicit attitudes may be the better predictors. This study is the first to examine the relations among parents’ implicit attitudes about smoking, parental smoking cessation, and adolescent smoking. The above discussion illustrates the potential importance of studying parental smoking cessation and its correlates in order to understand the processes underlying adolescents’ smoking behavior. However, there are also practical implications for intervention. Treatment programs might use parental influence on children as a source of motivation to support parents’ cessation efforts. These programs might also help parents discuss their own smoking cessation with their children in ways that foster the children’s nonsmoking. Thus, for both theoretical and practical reasons, it is important for researchers to ask whether parental smoking cessation is associated with a lowered likelihood of adolescent smoking and, if so, whether smoking-specific parenting behaviors and attitudes might mediate this effect. These are the goals of this study.

Method Participants Participants were adolescents ages 10 to 17 and their parents, at least one of whom was also in our longitudinal smoking survey (Chassin et al., 1984, 2000). All 1999 survey respondents who reported that they had a child between ages 10 and 18 were invited to participate in this family study (N = 861), and 556 (65%) accepted. Parents who refused were more likely to be smokers (45.0% of refusers vs. 32.7% of participants) and less likely to have attended college (21.8% of refusers vs. 38.8% of participants, ps < .05), suggesting some caution in generalization. Some families with a child over the age of 18 were recruited for another study. Informed parental consent and child assent were obtained at the time of the study, and the protocol was approved by the Indiana University Institutional Review Board.

487

Families were excluded if the child was 18 years old or older (n = 14), parental smoking cessation occurred within the past 6 months so that it might be unstable (n = 23), or the child did not live at least part time with two custodial parents (n = 73, final n for analysis = 446). (We did not consider legal custody, but were interested in environmental exposure to parental figures. Adolescents who lived full time with a single parent and had no contact with another parent figure [either because the other parent had died or had no contact] had to be eliminated because we could not consider the impact of their custodial mothers’ and fathers’ smoking status [and their interactions] on adolescent smoking.) In this final sample, 51% of the adolescents were female, 55% lived full time with two biological parents, and their average age was 12.8 years. Custodial parents’ age averaged 34.0 for mothers and 36.1 years for fathers. Because the sample was 98% non-Hispanic Caucasian, ethnic differences were not examined. Procedure Families living in the vicinity attended a laboratory session, and the others were interviewed in their homes by an interviewer from a nearby university. For both parents and adolescents (independently), the laboratory session began with a bioassay to validate self-reported smoking (using a MicroCO to measure carbon monoxide in expired air). Then, both parents and adolescents (independently) did a computer task (described below) to measure implicit attitudes toward smoking. Finally, parents completed a questionnaire (in separate rooms). In a different room, an interviewer read the questionnaire items aloud to the child, who responded on his or her own copy. All participants were assured of confidentiality. Measures Smoking Status. Adolescents reported whether they had ever smoked and their current smoking frequency from less than monthly to daily. Parents indicated their lifetime smoking status as follows: “never smoked, not even a single puff,” “smoked one or two, ‘just to try’, but not in the past month,” “no longer smoke, but used to be a regular smoker,” “smoke, but no more than once a month,” “smoke, but no more than once a week,” “smoke, but no more than once a day,” and “smoke more than once a day.” Parents and adolescents also reported the time since their last cigarette from within the past hour to more than 5 years

488

ago. Eighty percent of ex-smoking parents reported smoking their last cigarette at least 2 years ago. The bioassay supported the validity of these selfreports. Those who did and did not self-report smoking in the past 4 hours were compared in their bioassay readings as a smoker (i.e., CO levels greater than 10 parts per million; cf. Wald, Idle, Boreham, & Bailey, 1981). The resulting kappas ranged from .73 for adolescents to .91 for mothers (all ps < .0001). When a parent was not interviewed (4% of custodial mothers; 27% of the custodial fathers), that parent’s smoking was assessed by spousal reports. Parents were categorized as either nonsmokers (never smokers or triers), ex-smokers (used to smoke regularly but had not smoked for at least 6 months), or current smokers (smoke at least monthly). Given the young age of the adolescents, we used their lifetime smoking (dichotomized) as the outcome. Its prevalence was slightly lower than national levels for middle school participants (21.7% here vs. 29.3% nationally), but comparable to national levels for high school age participants (63.6% in this sample vs. 63.5% nationally; Centers for Disease Control and Prevention [CDC], 2000). Smoking-Specific Parenting: Normative Beliefs, Legitimacy to Regulate Adolescent Smoking, and Antismoking Behaviors. Parents reported their normative beliefs about their adolescents’ smoking, and adolescents reported their perceptions of their mothers’ and fathers’ beliefs. There were three items (e.g., parent’s value on the adolescent’s nonsmoking) with higher values indicating more antismoking normative beliefs. There was good internal consistency for adolescent report (α = .77 and .81 for perceptions of their mothers and fathers) but less for parental self-report (α = .54 and .60 for mothers and fathers). Parents and adolescents reported their perceptions of parental legitimacy to regulate the adolescent’s smoking (six items: e.g., “My mom experimented with smoking as a teenager, so she doesn’t have the right to try to stop me from smoking,” “It’s part of my personal choice whether or not to smoke, rather than something that my mom [dad] ought to control,” scored with higher values indicating greater legitimacy). Because this construct has never been tested, the items were created for this study. However, they showed good internal consistency (α = .85 for fathers, .86 for mothers, and .89 and .82 for adolescents’ ratings of their fathers and mothers). Finally, parents and adolescents reported the extent to which parents would react to the adolescent’s


smoking with antismoking behavior (eight items: e.g., “take away privileges,” with higher values indicating more antismoking behaviors). These items predicted adolescent smoking in an earlier study (Chassin et al., 1998) and showed good internal consistency in the current data (α = .83 for both mother and father-self-report, and .88 and .90 for adolescent report of mother and father). For all variables, parent self-report and children’s reports were significantly but modestly correlated (r s from .15 for father’s legitimacy to .32 for mother’s normative beliefs, all ps < .05). Attitudes About Smoking: Implicit and Explicit Attitudes Parents’ and adolescents’ implicit attitudes were measured using Greenwald et al.’s (1998) Implicit Association Test (IAT) procedure. The IAT uses a dual categorization task in which one of the categorizations involves a discrimination between positive and negative words, and the other involves a discrimination between two stimuli (here, smoking-related pictures vs. neutral shapes). In one case, the attitude object (e.g., cigarette) shares a response key with positive words, while the other (e.g., shapes) shares a response key with negative words. In the other case, these pairings are reversed. The difference in latency to respond indicates attitudes toward cigarettes. That is, if it is more difficult to use the same response key for cigarette images and positive words, then implicit attitudes toward cigarettes are negative. There were five phases, with on-screen instructions and a short practice before each one. In the first phase, positive and negative adjectives were shown. Participants responded as quickly as possible by using their right hand to press the “5” key on the number pad (if the word was good) or their left hand to press the “a” key (if the word was bad). The word remained on the screen until the participant responded. There were 16 trials with an intertrial interval of 250 ms. The second phase asked participants to categorize pictures as either a smoking stimulus or a shape, and the pictures remained on the screen until the participant responded. There were 32 trials with an intertrial interval of 250 ms. The third phase was a random presentation of both the words and the pictures. Participants categorized each picture and responded to each word with the same response keys they had practiced earlier. The stimulus remained on the


screen until the participant responded. There were 33 trials with an intertrial interval of 250 ms. The fourth phase was identical to phase two (categorizing only pictures) except that the right and left hand responses were reversed. The fifth phase was another combination task that was identical to phase three except that the right and left hand responses were reversed for categorizing the pictures. The measure of implicit attitudes was a difference score calculated by subtracting the mean latency score for compatible trials (i.e., trials in which the smoking pictures and “bad” adjectives shared the same response key) from the mean latency score for incompatible trials (i.e., trials in which the categorization of smoking pictures and “good” adjectives shared the same response key). Thus, higher scores reflect more negative implicit attitudes toward smoking. Following Greenwald et al.’s (1998) recommendations, we dropped the first two trials of each task from scoring to allow for adaptation to the task, and extreme latencies (under 300 ms and over 3,000) were recoded to equal 300 and 3,000 ms respectively. Participants with average error rates greater than 25% were eliminated (n = 18 adolescents and 7 parents). Children’s implicit attitudes were significantly (although weakly) correlated with mother’s (r = .15, p < .05) but not father’s (r = .06) implicit attitudes. There is now substantial evidence for the IAT’s reliability, as well as convergent and discriminant validity (Cunningham, Preacher, & Banaji, 2001; Greenwald & Nosek, in press). Test retest reliability has averaged above .6, and split-half reliabilities are about .90 (Bosson, Swann, & Pennebaker, 2000; Greenwald & Farnham, 2000). Construct validity has been demonstrated by showing that the IAT correlates in expected ways both with other implicit measures of similar constructs (Cunningham et al., 2001) and with group membership classification and measures of individual differences (Greenwald et al., 1998). Explicit global attitudes toward smoking were measured with five items asking if smoking was “nice,” “pleasant,” “fun,” and “good” along with a rating of attitude from “very negative” to “very positive,” all with higher values indicating more positive attitudes (Ajzen & Fishbein, 1980). These items prospectively predicted adolescent smoking (Chassin et al., 1984) and had high internal consistency in this study (α = .89 for children and fathers and .91 for mothers). Children’s and parent’s explicit attitudes were significantly (although modestly) correlated (rs = .12 for fathers and .21 for mothers, both ps < .05).

489

Results Relations Between Parental Smoking and Adolescent Smoking We tested whether parental smoking cessation was associated with lowered rates of child smoking using hierarchical logistic regression to predict adolescent smoking (dichotomized) from mothers’ and fathers’ smoking (non/ex/current; two dummy-coded variables for each) and the interaction between mothers’ and fathers’ smoking (four variables). Child age, family structure (child living full time with two biological parents vs. any other), and parent education were entered in the first block as covariates. In preliminary analyses, we also tested the effects of gender, but because it was unrelated to parent smoking status and to adolescent smoking, we did not pursue its effects. All covariates had significant effects such that older children, those with less educated parents, and those who did not live full time with two biological parents were more likely to have smoked (Wald χ2 s from 15.6 to 61.2, all ps < .001). The block of two maternal smoking variables added significant explained variance above and beyond the covariates, (χ2 [2] = 8.32, p < .02, as did the paternal smoking variables (χ2 [2] = 9.64, p < .008). Finally, the block of variables representing the interaction between maternal and paternal smoking was marginally significant above and beyond both the covariates and main effects of parent smoking (χ2 [4] = 7.97, p < .09). The main effects of maternal and paternal smoking and their interaction produced a 6% change in the Negelkerke R2 over and above the covariates. Table I presents the prevalence of child smoking for the combinations of maternal and paternal smoking of interest (i.e., families with one currently smoking and one never smoking parent are omitted). We compared the prevalence within each cell to the most low-risk (two nonsmoking parents) and the most high-risk (two smoking parents) combinations above and beyond the effects of the covariates. Compared to those with two nonsmoking parents, children with two smoking parents were significantly more likely to smoke (Wald χ2 [1] = 15.01, p < .001, odds ratio [OR] = 4.19), and those with a smoking mother and an ex-smoking father (Wald χ2 [1] = 3.57, p < .06, OR = 4.21) were marginally more likely to smoke. We next compared each cell to those with two smoking parents. Those with two smoking parents were significantly more likely to smoke than were those with a

490

Table I.


Adjusted Group Means for Variables That Showed Interactions Between Maternal and Paternal Smoking

Measure % adolescents with any

Both nona

Mom non, dad exb

14.1a

25.0ad

Mom ex, dad nonc

Both parents exd

Mom smoker, dad exe

Mom ex, dad smokerf

Both smokersg

37.0bcd

21.4ac

50.0bc

38.1ac

53.9b

smoking h Mother’s parenting: Child report Normative beliefs

4.87a

4.87a

4.83a

4.82a

4.64ab

4.78a

4.49b

Legitimacy

3.94a

3.93ab

3.95ab

3.98ab

3.27b

3.66ab

3.76ab

Antismoking behavior

4.40a

4.31a

4.09ab

4.41a

4.10ab

3.98ab

3.95b

4.35a

4.38a

4.39a

4.30ab

3.93b

4.25ab

4.17ab

Normative beliefs

4.74a

4.77ac

4.60ab

4.71ab

4.33b

4.30bc

4.44b

Legitimacy

4.01a

4.03ab

4.01ab

3.80ab

3.49b

3.88ab

3.79ab

Antismoking behavior

4.28a

4.18ac

3.97ab

4.37a

3.80ab

3.67bc

3.84b

4.11ac

4.25ac

4.12acd

4.40a

3.66bd

3.71b

3.98bc

Father’s implicit attitude

282.89a

254.20a

306.80a

174.91b

Father’s explicit attitude

1.38a

1.43a

Mother’s parenting: Self-report Legitimacy Father’s parenting: Child report

Father’s parenting: Self-report Antismoking behavior Attitudes 266.04ab 1.20a

349.23a

374.45a

1.54ac

2.00c

2.62bc

2.74b

Numbers vary because of missing data. Means in the same row that do not share subscripts differ at p < .05 or show odds ratios > 2.5. Higher values indicate intolerance of adolescent smoking, greater legitimacy to regulate adolescent smoking, more antismoking parenting behavior, more negative implicit attitudes, and less negative explicit attitudes. a Numbers ranged from 104 to 156. b Numbers ranged from 33 to 52. c Numbers ranged from 16 to 27. d Numbers ranged from 9 to 14. e Numbers ranged from 5 to 12. f Numbers ranged from 11 to 21. g Numbers ranged from 54 to 99. h Percentages represent observed prevalences of smoking.

nonsmoking mother and an ex-smoking father (Wald χ2 [1] = 7.42, p < .01, OR = 3.33) and were marginally more likely to smoke than were those with either two ex-smoking parents or an ex-smoking mother and a smoking father (Wald χ2 s[1] = 3.06 and 2.78, respectively, both ps < .10, ORs of 3.85 and 2.63, respectively). Note that those with a smoking mother and ex-smoking father did not differ from those with two smoking parents. Relations Between Parental Smoking and Hypothesized Mediators We tested whether parental smoking was related to the hypothesized mediators using a series of 3 (mothers’ smoking: non/ex/current) by 3 (fathers’ smoking: non/ex/current) analyses of covariance (ANCOVAs with child age, family structure, and parent education as covariates). The sample sizes in these analyses varied because we had fewer participating fathers than mothers and a smaller sample in the laboratory that completed the implicit attitude task. Thus, the power to detect a small interaction effect in these analyses was not optimal and varied from .65 to .41.

Accordingly, we probed all of the interactions that were conventionally significant (p < .05) or that accounted for at least 1.5 % of unique variance (above and beyond the covariates and the main effects of maternal and paternal smoking). Results of these ANCOVAs are shown in Table II, and means and pairwise comparisons for the interactions that were probed are shown in Table I. Moreover, for each interaction that was probed, we performed three additional planned contrasts to test the impact of parent cessation when the other parent still smoked (aggregating across some of the cells in Table I). We compared those with one ex-smoking parent and one smoking parent to (1) those with two nonsmoking parents, (2) those with two smoking parents, and (3) those with one ex-smoking parent and one nonsmoking parent (with significant contrasts described in the text). Mothers’ Parenting. For child reports of mothers’ normative beliefs, there was an interaction of maternal and paternal smoking (see Tables I and II). Children with two smoking parents perceived their mothers as most accepting of their smoking, and this group significantly differed from all others except from


Table II.

491

Relations of Parental Smoking to the Hypothesized Mediators (ANCOVAs) Mother smoking

Hypothesized mediator

F value

Interaction of mother and father smoking

Father smoking η2

F value

η2

F value

η2

n

Mother’s parenting: Child report Normative beliefs

6.77***

.031

2.09

.010

2.89*

.026

441

Legitimacy

0.90

.004

3.23*

.015

2.02t

.018

441

Antismoking behaviors

3.78*

.017

2.29t

.011

1.81

.017

441

Mother’s parenting: Self-report Normative beliefs

12.80***

.058

0.95

.005

0.52

.005

428

Legitimacy

4.07*

.019

0.22

.001

1.98t

.019

428


7.47***

.035

0.15

.001

0.55

.005

428

Father’s parenting: Child report Normative beliefs

1.19

.006

6.45**

.030

3.88**

.035

436

Legitimacy

0.48

.002

2.93t

.014

2.14t

.020

436


3.99*

.018

2.70t

.013

2.54*

.023

436

Father’s parenting: Self-report Normative beliefs

1.11

.006

2.30

.013

1.23

.014

363

Legitimacy

0.24

.001

1.92

.011

1.23

.014

363


3.03*

.017

0.91

.005

2.02t

.022

363

Child

1.29

.007

2.42t

.013

0.80

.009

367

Mother

3.82*

.022

1.01

.006

0.69

.008

353

Father

0.58

.004

1.74

.013

2.56*

.038

268

1.06

.005

1.21

.006

1.42

.013

433

187.90***

.476

0.06

.000

1.10

.010

426

.031

108.04***

.387

1.31

.015

354

Implicit attitudes

Explicit attitudes Child Mother Father

5.49**

Degrees of freedom range: 2,256 to 2,429 for main effects and 4,256 to 4,429 for interactions. Sample sizes vary due to missing data. t p < .10. *p < .05. **p < .01. ***p < .001.

children with a smoking mother and an ex-smoking father. None of the additional planned contrasts was significant. For child reports of mothers’ legitimacy, there was a marginally significant interaction of maternal and paternal smoking that accounted for 1.8% of unique variance (see Tables I and II). Children with a smoking mother and an ex-smoking father viewed their mothers as having the least legitimacy to regulate their smoking, and this group differed from children of two nonsmoking parents. Moreover, children with one ex-smoking and one smoking parent viewed their mothers as having significantly less legitimacy than did those with two nonsmoking parents (p < .01, η2 = .016) and those with an ex-smoking and a nonsmoking parent (p < .02, η2 = .014), and they did not differ from those with two smoking parents. Finally, for children’s reports of their mothers’ antismoking behaviors, there was an interaction of maternal and paternal smoking that accounted for 1.7% of unique variance (see Tables I and II). Children of

two ex-smoking parents and two-nonsmoking parents saw their parents as most antismoking, significantly more so than did children of two smoking parents. However, those with one ex-smoking and one smoking parent did not significantly differ from those with two smoking parents. No additional planned contrasts were significant. When mothers’ self-reports were considered, there was a significant effect of mothers’ smoking for all three variables. In each case, smoking mothers were the most distinct, and most permissive of smoking (Madj = 4.61 for normative beliefs, 4.02 for perceived legitimacy, and 3.83 for antismoking behaviors) compared to ex-smoking and nonsmoking mothers (Madj = 4.94 and 4.98 for normative beliefs, 4.31 and 4.28 for legitimacy, and 4.18 and 4.21 for antismoking parenting behaviors). For mother-reported legitimacy, there was also a marginally significant interaction of maternal and paternal smoking that accounted for 1.9% of unique

492

variance (see Tables I and II). Smoking mothers in families with ex-smoking fathers self-reported the least legitimacy (paralleling the child’s perceptions). Moreover, mothers in families with one ex-smoking parent and one smoking parent reported significantly less legitimacy than did mothers in families with: two nonsmoking parents ( p < .05, η2 = .01) and b) one ex-smoking and one nonsmoking parent ( p < .04, η2 = .01), and they did not differ from families with two smoking parents. Fathers’ Parenting. For children’s reports, there were interactions between mothers’ and fathers’ smoking for each of the three variables accounting for between 2% and 3.5% of unique variance (see Tables I and II). For normative beliefs, children with one smoking parent and one ex-smoking parent viewed their fathers as more tolerant than did children in families with two nonsmoking parents ( p < .002, η2 = .02) and one ex-smoking and one nonsmoking parent ( p < .02, η2 = .015) and as tolerant as children with two smoking parents. Legitimacy showed a similar pattern. Children with an ex-smoking father and a smoking mother perceived their fathers as having the least legitimacy to regulate their smoking. Finally, children in families with smoking fathers and ex-smoking mothers viewed their fathers as showing the least antismoking behavior. Children in families with one ex-smoking parent and one smoking parent viewed their fathers as less antismoking than did children with two nonsmoking parents ( p < .001, η2 = .02), and they did not significantly differ from those with two smoking parents. Fathers’ self-reports produced a marginally significant interaction for antismoking behaviors that accounted for 2.2% of unique variance (see Tables I and II). Fathers in families with one ex-smoking parent and one smoking parent self-reported the least antismoking parenting behavior (paralleling the child’s perceptions). These fathers were less antismoking than were fathers in families with two nonsmoking parents ( p < .02, η2 = .017) and one ex-smoking and one nonsmoking parent ( p < .006, η2 = .02), and they were only marginally different from those with two smoking parents ( p < .09, η2 < .01). Implicit Attitudes Toward Smoking. Children’s implicit attitudes showed no significant unique relations to parental smoking. However, smoking mothers had more positive implicit attitudes (Madj = 226.98) than did either nonsmoking mothers (Madj = 298.50) or ex-smoking mothers (Madj = 292.50). Fathers’ implicit attitudes showed a significant interaction between mothers’ and fathers’ smoking (see Tables I and


II). Fathers in families with two smoking parents had the most positive implicit attitudes, and they significantly differed from all other groups except those with ex-smoking mothers and nonsmoking fathers. Thus, even if the father smoked, when the mother was an ex-smoker, his implicit attitudes were more negative than when the mother was a current smoker. Explicit Attitudes Toward Smoking. Children’s explicit attitudes showed no significant unique relations to parent smoking. Mothers’ explicit attitudes showed a main effect of maternal smoking such that smoking mothers had the most positive attitudes (Madj = 2.51), compared to nonsmoking mothers (Madj = 1.18) and ex-smoking mothers (Madj = 1.32). Finally, fathers’ explicit attitudes showed an interaction between maternal and paternal smoking that explained 1.5% of unique variance. Fathers in families with two smoking parents were the most positive and differed from all groups except those with smoking fathers and ex-smoking mothers (see Table I). Fathers in families with one ex-smoking and one smoking parent were more positive toward smoking than were fathers in families with two nonsmoking parents (η2 = .115, p < .001), more positive than those with one ex-smoking parent and one nonsmoking parent (η2 = .118, p < .001), but more negative than fathers in families with two smoking parents (η2 = .025, p < .003). Relations Between the Hypothesized Mediators and Adolescent Smoking Relations between the hypothesized mediators and adolescent smoking were tested in 10 logistic regression models (all including child age, family structure, parental education, maternal smoking, paternal smoking, and the interaction of maternal and paternal smoking). As noted earlier, older adolescents, those who did not live full time with two biological parents, and those who had less educated parents were more likely to have smoked (all ps < .05). Moreover, significant parent smoking effects were never entirely eliminated by adding the mediators. The hypothesized mediators were entered in blocks in separate models as follows: parenting models (normative beliefs, perceived legitimacy, and antismoking behaviors; separate models for child report about mother, mother’s self-report, child report about father, and father’s self-report), implicit attitudes (separate models for child, mother, and father), and explicit global attitudes (separate models for child, mother, and father). Standardized scores for the me-


diators were used so that their associated ORs could be interpreted, as the increased odds of adolescent smoking given a one standard deviation change in the mediator. Mothers’ Parenting. Child reports of mothers’ antismoking behaviors (but not normative beliefs or legitimacy) had a unique effect on adolescent smoking (Wald χ2[1] = 13.94, p < .001, OR = 2.15, such that those who viewed their mothers’ parenting behaviors as less antismoking were more likely to have smoked. When the model was estimated with mothers’ selfreports, there was a significant effect of normative beliefs, but not legitimacy or behaviors (Wald χ2[1] = 4.89, p < .03, OR = 1.37), such that mothers who reported less negative beliefs about their child’s smoking were more likely to have a child who had smoked. Fathers’ Parenting. For child reports of fathers’ parenting, there was a significant effect of antismoking behaviors, but not legitimacy or normative beliefs (Wald χ2[1] = 15.57, p < .001, OR = 2.33), such that adolescents who viewed their fathers’ parenting behaviors as less antismoking were more likely to have smoked. Considering fathers’ self-reports, there were no significant unique effects of the parenting mediators. Implicit Attitudes About Smoking. Neither children’s nor fathers’ implicit attitudes had unique effects on adolescent smoking. Mothers with more positive implicit attitudes had children who were more likely to have smoked (Wald χ2[1] = 4.95, p < .03, OR = 1.41). Explicit Attitudes About Smoking. Children’s attitudes had a significant unique relation to their smoking (Wald χ2[1] = 29.35, p < .001, OR = 2.67), such that those with more positive attitudes were more likely to have smoked. Mothers’ and fathers’ self-reported explicit attitudes did not uniquely predict adolescent smoking. Tests of Mediated Effects Four variables met Baron and Kenny’s (1986) preconditions for mediation (i.e., significant relations with parent smoking and significant unique relations with child smoking over and above parent smoking). For these variables (child reports of mothers’ and fathers’ antismoking parenting behaviors, mothers’ implicit attitudes, and mothers’ self-reported normative beliefs), we used MPLUS software (Muthen & Muthen, 1998) to obtain the relevant path estimates for mediated effects. Multinomial predictors were entered using weighted contrast coding. The standard error of the mediated effect was calculated by extending the

493

multivariate delta method proposed by MacKinnon (2000; MacKinnon, personal communication). In these models, the tests of the mediated effects represent the combined main effects of maternal and paternal smoking and their interactions.1 There was no significant mediated effect for mothers’ normative beliefs. However, there was a marginally significant mediated effect for mother’s implicit attitude (path estimate/standard error = 1.88, p < .10) and significant effects for the child’s reports of parenting behavior (path estimate/standard error = –34.34 for mother’s and –50.22 for father’s parenting, both ps < .001).

Discussion This study asked whether parental smoking cessation was associated with lowered levels of adolescent smoking and whether smoking-specific parenting behaviors and attitudes (both implicit and explicit) were plausible candidates to mediate such effects. Results showed that parent smoking status had a moderate-sized effect on adolescent smoking. Moreover, this effect was found above and beyond both the effects of child age and of demographic risk factors, thus representing a stringent test of parent smoking influences. Moreover, parent smoking cessation was indeed associated with reduced prevalence of adolescent smoking. Adolescents with an ex-smoking parent had lowered prevalence of smoking except when their other custodial parent (particularly the mother) was a current smoker. Although causal relations cannot be identified with our correlational design, attempts to manipulate parental smoking (i.e., in cessation interventions) could test the impact of parents’ successful quitting on their adolescent children to determine whether parental smoking treatment can function as a form of preventive intervention for the next generation. Interestingly, the beneficial effect of parental smoking cessation was largely negated if the other parent continued to smoke, particularly if that smok1 Weighted contrast coding was necessary because the available formula for the standard error does not account for indirect paths passing through correlations between coded variables. Weighted contrast coding forced all indirect effects to be accounted for in the direct paths between the coded variables and the mediators and the path from the mediator to the outcome, thus providing an appropriate estimate of the total mediated effect. However, the weighted contrast coding accounts for both the interaction and main effect terms, so that tests for the mediated effect combine both the main effects of maternal and paternal smoking and their interaction. Further research is required to develop a method for decomposing this total mediated effect into the mediated effect from the main effects and from the interaction term.

494

ing parent was the mother. Smoking cessation programs already recognize that the presence of a smoking spouse raises risk for an individual’s relapse after quitting. Our data suggest that the presence of a currently smoking spouse also threatens the potential benefits for the next generation. We also asked whether smoking-specific parenting behaviors could mediate this effect. We found that parent smoking cessation was related to smoking-specific parenting variables that have been shown in other studies to reduce the risk of adolescent smoking. Namely, ex-smoking mothers viewed themselves as quite intolerant of their child’s smoking, and both ex-smoking mothers and fathers viewed themselves as antismoking in their parenting (except for ex-smoking fathers in families with a currently smoking mother). However, the extent to which the children perceived such parental intolerance of their smoking and these antismoking parenting behaviors was undermined for children of exsmoking parents when the other parent continued to smoke. Parents in these circumstances may find it difficult to frame credible antismoking messages for their children without somehow implying criticism of the other parent. A lack of consistency between parents’ behaviors and their messages (and between the two parents) may create a mixed signal for these adolescents that undermines the benefits of parental cessation. These findings are consistent with the principles of social learning theory in that consistent models should be the most powerful agents of social learning, whereas inconsistent models (such as one smoking parent and one ex-smoking parent) should be less effective in transmitting parental messages. These findings highlight the importance of children’s perceptions of their parents’ antismoking behaviors because these variables were not only related to parent smoking but also showed unique relations to adolescent smoking (above and beyond parental smoking). In fact, our analyses showed that children’s perceptions of their parents’ antismoking behaviors partially mediated the effects of parent smoking on adolescent smoking. This finding replicates and extends earlier work that simply compared smoking and nonsmoking parents without considering the effects of parental smoking cessation (e.g., Chassin et al., 1998). Parents’ perceived legitimacy to regulate smoking was also related to parent smoking. Contrary to prediction, ex-smoking parents did not necessarily view themselves as having less legitimate authority to reg-


ulate their child’s smoking just because they themselves had smoked in the past. However, adolescents in families with one ex-smoking parent and one current smoking parent did view their parents as having this lowered legitimacy, particularly when the mother smoked, (paralleling findings for parental normative beliefs and antismoking behaviors already described). Perceived legitimacy has not been studied in the context of adolescent smoking, and it did not uniquely relate to adolescent smoking in these data. However, lowered parental legitimacy is still of concern because adolescents should be less likely to internalize and comply with parental directives that they view as not based in legitimate authority (Grusec & Goodnow, 1994). Thus, this variable warrants further study. The importance of attitudinal variables in our data was less clear because children’s attitudes (both implicit and explicit) were unrelated to parental smoking. Thus, although (as typically found) children’s explicit attitudes to smoking were correlated with their smoking behavior, the children’s attitudes do not appear to be shaped uniquely by their parents’ smoking. Interestingly, however, parents’ implicit attitudes toward smoking were related to parental smoking and (for mothers) uniquely related to the adolescent’s smoking as well. This study is the first to demonstrate that mothers’ implicit attitudes toward smoking relate to their child’s smoking and, in fact, may be better predictors than are explicit measures of mothers’ attitudes. In fact, mothers’ implicit attitudes showed a marginally significant mediational pathway from parent smoking to child smoking. Although beyond the scope of this study, implicit attitudes may reflect affective messages that can interact with parenting strategies that are under more conscious control. In any case, the unique relation of mothers’ implicit attitudes to adolescent smoking is a novel and potentially important finding that warrants further study. In relating parenting and attitudes to the adolescent’s smoking behavior, many of the child-reported variables, fewer of the mother-reported variables, and none of the father-reported variables were unique predictors (above and beyond child age, family structure, parent education, and parent smoking). This may reflect meaningful and important reporter effects. That is, parents’ attitudes or their beliefs that they provide antismoking parenting might be ineffective unless their adolescents also internalize these perceptions (Grusec & Goodnow, 1994), and our


modest correlations between parents’ and adolescents’ perceptions suggest less than strong transmission from parent to child. Thus, adolescents’ perceptions might be the most important in influencing their smoking decisions. However, methodological considerations limit the certainty with which we can interpret any lack of relation between the hypothesized mediators and adolescent smoking. Because this young sample has not passed through the age of risk for smoking initiation, we cannot distinguish true nonsmokers from those who will initiate smoking in the future. Stronger effects of parent-reported variables might be found in predictions of longerterm outcomes. Although this study extended previous research by examining potential mediators of parent cessation effects, by including both implicit and explicit attitudes and both parent and adolescent reports, and by validating self-reported smoking status, it does have limitations. The young age of the subjects and the cross-sectional design dictate caution in drawing conclusions about the influence of attitudinal and parenting variables on adolescent smoking. Moreover, although the sample size was relatively large for a laboratory-based family study, our statistical power to detect small effects (particularly pairwise differences between family types) was not optimal. It would be particularly interesting to examine larger samples of families with one ex-smoking and one currently smoking parent. Finally, our data do not allow us to examine genetic mechanisms underlying parent smoking effects. In short, these findings suggest that parental smoking cessation may have potential benefits for

495

adolescents in terms of a lower prevalence of smoking, but that these benefits may be undermined when one of the parents (particularly the mother) continues to smoke. The relation between parental smoking and adolescent smoking may be partly due to adolescents’ perceptions of their parents’ antismoking behaviors, and inconsistencies between the two parents in their smoking behavior may undermine parents’ abilities to frame credible messages. Future studies of the role of parents’ implicit attitudes toward smoking and of parents’ perceived legitimacy to regulate adolescent smoking are also warranted, particularly with larger samples of ex-smoking parents and longitudinal designs.

Acknowledgments This research was supported by Grants DA13555 and K05DA00492 from the National Institute on Drug Abuse and HD13449 from the National Institute of Child Health and Human Development. We thank Alma Taubensee and William Rodawalt for coordinating data collection and David MacKinnon, Antonio Morgan-Lopez, and Linda Muthen for consultation on testing mediated effects. Portions of these data were presented at the 11th World Congress on Tobacco OR Health, Chicago, June 2000, in a symposium sponsored by GlaxoSmithKline and Pharmacia, and those analyses were partially supported by an unrestricted grant from GlaxoSmithKline. Received December 21, 2000; revisions received May 30, 2001; accepted August 15, 2001

References Ajzen, I., & Fishbein, M. (1980). Understanding attitudes and predicting social behavior. Englewood Cliffs, NJ: PrenticeHall. Bailey, S. L., Ennett, S. T., & Ringwalt, C. L. (1993). Potential mediators, moderators, or independent effects in the relationship between parents’ former and current cigarette use and their children’s cigarette use. Addictive Behaviors, 18, 601–621. Baron, R., & Kenny, D. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173–1182. Bauman, K. E., Foshee, V. A., Linzer, M. A., & Koch, G. G. (1990). Effect of parental smoking classification on the

association between parental and adolescent smoking. Addictive Behaviors, 9, 413–422. Boomsma, D., Koopmanns, J., van Doornen, L., & Orlebeke, J. (1994). Genetic and social influences on starting to smoke: A study of Dutch adolescent twins and their parents. Addiction, 89, 219–226. Bosson, J. K., Swann, W. B., & Pennebaker, J. W. (2000). Stalking the perfect measure of self-esteem: The blind men and the elephant revisited. Journal of Personality and Social Psychology, 79, 631–643. Bricker, J., Leroux, B., Peterson, A., Kealey, K., Sarason, I., Anderson, M. R., & Marek, P. (2001). A nine-year prospective analysis of the relationship between parent smoking cessation and children’s daily smoking. Presented at the Annual

496

Meeting of the Society for Behavioral Medicine, April, Seattle, WA. Centers for Disease Control and Prevention (2000). Youth tobacco surveillance—United States, 1998–1999. Morbidity and Mortality Weekly Reports, 49(SS10), 1–93. Chassin, L., Presson, C. C., Pitts, S. C., & Sherman, S. J. (2000). The natural history of cigarette smoking from adolescence to adulthood in a midwestern community sample: Multiple trajectories and their psychosocial correlates. Health Psychology, 19, 223–231. Chassin, L., Presson, C. C., Rose, J., & Sherman, S. J. (1998). The inter-generational transmission of beliefs about smoking and adolescent smoking. Psychology of Addictive Behaviors, 12, 206–213. Chassin, L., Presson, C. C., Sherman, S. J., Corty, E., & Olshavsky, R. W. (1984). Predicting the onset of cigarette smoking in adolescents: A longitudinal study. Journal of Applied Social Psychology, 14, 224–243. Chassin, L., Presson, C. C., Todd, M., Rose, J., & Sherman, S. J. (1998). Maternal socialization of adolescent smoking: The intergenerational transmission of parenting and smoking. Developmental Psychology, 34, 1189–1201. Conrad, K., Flay, B., & Hill, D. (1992). Why children start smoking cigarettes: Predictors of onset. British Journal of Addiction, 87, 1711–1724. Cunningham, W. A., Preacher, K. J., & Banaji, M. R. (2001). Implicit attitude measures: Consistency, stability, and convergent validity. Psychological Science, 12, 163–170. Farkas, A., Distefan, J., Choi, W., Gilpin, E., & Pierce, J. (1999). Does parental smoking cessation discourage adolescent smoking? Preventive Medicine, 28, 213–218. Flay, B. R., Hu, F., & Richardson, J. (1998). Psychosocial predictors of different stages of cigarette smoking among high school students. Preventive Medicine, 27, 9–18. Flay, B. R., Petraitis, J., & Hu, F. (1999). Psychosocial risk and protective factors for adolescent tobacco use. Nicotine and Tobacco Research, 1(Suppl.), 59–66.


Greenwald, A. G., & Farnham, S. D. (2000). Using the Implicit Association Test to measure self-esteem and selfconcept. Journal of Personality and Social Psychology, 79, 1022–1038. Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998). Measuring individual differences in implicit cognition: The Implicit Association Test. Journal of Personality and Social Psychology, 74, 1464–1480. Greenwald, A. G., & Nosek, B. A. (in press). Health of the Implicit Association Test at age 3. Zeitschrift fur Experimentelle Psychologie. Grusec, J. E., & Goodnow, J. J. (1994). Impact of parental discipline methods on the child’s internalization of values: A reconceptualization of the current points of view. Developmental Psychology, 30, 4–19. Heath, A., & Madden, P. (1995). Genetic influences on smoking behavior. In J. R. Turner, L. Cardon, & J. K. Hewitt (Eds.), Behavioral genetic approaches to behavioral medicine (pp. 45–65). New York: Plenum. Jackson, C., & Henriksen, L. (1997). Do as I say: Parent smoking, antismoking socialization, and smoking onset among children. Addictive Behaviors, 22, 107–114. MacKinnon, D. P. (2000). Contrasts in multiple mediator models. In J. Rose, L. Chassin, C. C. Presson, & S. J. Sherman (Eds.), Multivariate applications in substance use research: New methods for new questions (pp. 141–160). Mahweh, NJ: Lawrence Erlbaum. Muthén, L. K., & Muthén, B. (1998). Mplus user’s guide. Los Angeles, CA: Muthén & Muthén. U.S. Department of Health and Human Services. (1994). Preventing tobacco use among young people: A report of the Surgeon General. Washington, DC: U.S. Government Printing Office. Wald, N. J., Idle, M., Boreham, J., & Bailey, A. (1981). Carbon monoxide in breath in relation to smoking and carboxyhaemoglobin levels. Thorax, 36, 366–369.